As the demand for powerful AI models grows, so does the compute required to train them. This episode dives into Epoch AI’s findings on how the compute for frontier AI models has been growing at a staggering 4-5x per year. We’ll discuss the implications of this rapid increase for both the capabilities and accessibility of advanced AI systems, including potential bottlenecks in hardware and energy consumption. Join us as we explore the trajectory of compute growth and what it tells us about the future of AI development, from language models to multimodal and autonomous systems.

Download Link:
https://epochai.org/blog/training-compute-of-frontier-ai-models-grows-by-4-5x-per-year